========

SETUP

========

FILEPATH:

filepath = "https://raw.githubusercontent.com/shabanm2/Utqiagvik/main/Analysis_Ready_Data/"

PICK DATE RANGES:

years = c("2022", "2023")
seasons = c("Spring", "Summer","Fall")
sites = c("TNHA", "BEO", "SSMH")

# put season values for the season that has the start of the data
date_start = "2022-06-01" # data starts in June 2022 (YEAR-MO-DY) where day is always 01
# put season values for the season that has the last of the data
date_end = "2023-11-01" # data ends after November of 2023 (will get data up UNTIL date_end not after)

PICK OUTPUT:

scree = F # scree plot
eigen = T # eigenvectors and eigenvalues

Dates (for selecting date ranges; no need to edit)

spring_months = c("March", "April", "May")
summer_months = c("June","July","August")
fall_months = c("September", "October", "November")
winter_months = c("December", "January", "February")

spring_dates = data.frame(months=spring_months, start=c("-03-01","-04-01","-05-01"), end=c("-04-01","-05-01","-06-01"))

summer_dates = data.frame(months=summer_months, start=c("-06-01","-07-01","-08-01"), end=c("-07-01","-08-01","-09-01"))

fall_dates = data.frame(months=fall_months, start=c("-09-01","-10-01","-11-01"), end=c("-10-01","-11-01","-12-01"))

winter_dates = data.frame(months=winter_months, start=c("-12-01","-01-01","-02-01"), end=c("-01-01","-02-01","-03-01"))

all_dates = data.frame(matrix(nrow = 0, ncol = 4))
for(yur in years){
  # spring
  if("Spring" %in% seasons){
    for(i in c(1:3)){
      curdate = as.Date(paste0(yur, spring_dates[i, 2]))
      if(curdate >= date_start && curdate < date_end){
        all_dates <- rbind(all_dates, c(spring_dates[i,1], paste0(yur, spring_dates[i, 2]), paste0(yur, spring_dates[i, 3]),"Spring"))
      }
    }
  }
  if("Summer" %in% seasons){
    for(i in c(1:3)){
      curdate = as.Date(paste0(yur, summer_dates[i, 2]))
      if(curdate >= date_start && curdate < date_end){
        all_dates <- rbind(all_dates, c(summer_dates[i,1], paste0(yur, summer_dates[i, 2]), paste0(yur, summer_dates[i, 3]),"Summer"))
      }
    }
  }
  if("Fall" %in% seasons){
    for(i in c(1:3)){
      curdate = as.Date(paste0(yur, fall_dates[i, 2]))
      if(curdate >= date_start && curdate < date_end){
        all_dates <- rbind(all_dates, c(fall_dates[i,1], paste0(yur, fall_dates[i, 2]), paste0(yur, fall_dates[i, 3]),"Fall"))
      }
    }
  }
  if("Winter" %in% seasons){
    for(i in c(1:3)){
      curdate = as.Date(paste0(yur, winter_dates[i, 2]))
      if(curdate >= date_start && curdate < date_end){
        all_dates <- rbind(all_dates, c(winter_dates[i,1], paste0(yur, winter_dates[i, 2]), paste0(yur, winter_dates[i, 3]),"Winter"))
      }
    }
  }
  
}
colnames(all_dates) = c("months","start","end","szn")

Packages

library(dplyr)
library(lubridate)
library(tidyverse)

Load all data

daily <- read.csv(paste0(filepath, "daily_2022_2024.csv"))
daily <- daily %>% select(-X) %>% select(-X.1) # get rid of index columns
daily$Date <- as.POSIXct(daily$date, format="%Y-%m-%d") # format dates
daily$fullname[daily$site == "BEO"] <- "BEO-BASE"
daily <- daily %>% filter(fullname == "TNHA-SA" | fullname == "TNHA-SC" | fullname == "SSMH-SB" | fullname == "SSMH-SA" | fullname == "BEO-BASE") %>% select(-c(winddir, date)) %>% mutate(aspect = case_when(fullname == "TNHA-SC" | fullname == "SSMH-SB" ~ "North", fullname == "TNHA-SA" | fullname == "SSMH-SA" ~ "South", .default = "N/A")) %>% filter(grounddepth == 8) %>% filter(Date >= "2022-06-19") # create "aspect" column and filter for top depth of soil and start date of when we started collecting data
# note: the data before June 19, 2022 was estimated by our gap-filling script and should be disregarded due to extrapolation

=========================

DEFINE FUNCTIONS

=========================

Temporal Range: Season Vertical Spatial Range: 30-45 cm Horizontal Spatial Range: stations across site (TNHA, SSMH, BEO) –> Average Total Site –> North vs South (except for BEO)

Filter by Site and Join Tables

pick_site <- function(cursite){

  big_df <<- daily %>% filter(site == cursite)
  
  if(szn == "Winter"){
    big_df <<- big_df %>% select(-solar)
  }
  return(big_df)
}

Filter by Date Range

pick_dates <- function(datemin, datemax, big_df){
  pca_df <<- big_df %>% filter(Date >= datemin) %>% filter(Date < datemax)
  
  # get rid of NAs
  pca_df <<- na.omit(pca_df)
  pca_df <<- unique(pca_df)
  return(pca_df)
}

Calculate PCA

calc_pca <- function(pca_df){
  pca <<- prcomp(pca_df[,6:(ncol(pca_df)-2)], center=TRUE, scale.=TRUE)

  #take out variables
  sd <- pca$sdev
  loads <<- pca$rotation
  rownames(loads) <<- colnames(pca_df[6:(ncol(pca_df)-2)])
  scores <<- pca$x
  
  var <- sd^2
  varPercent <- var/sum(var) * 100
  
  return(list("pca"=pca, "loads"=loads))
}

Make Scree Plot

make_scree <- function(pca){
  sd <- pca$sdev
  
  var <- sd^2
  varPercent <- var/sum(var) * 100
  
  barplot(varPercent, xlab="PC", ylab="Percent Variance", names.arg=1:length(varPercent), 
          las=1, ylim=c(0, max(varPercent)), col="gray")
  abline(h=1/ncol(pca_df[5:ncol(pca_df)])*100, col="red")
}

Display Eigenvectors and Eigenvalues

make_eigen <- function(pca){
  eigenvectors <- pca$rotation
  print("Eigenvectors (Loadings):")
  print(eigenvectors)
  
  print("Loadings Cutoff:")
  sqrt(1/ncol(pca_df[5:ncol(pca_df)])) # cutoff for "important" loadings
  
  # Access the eigenvalues (variances of the principal components)
  eigenvalues <- (pca$sdev)^2
  print("Eigenvalues:")
  print(eigenvalues)
}

===============

PCA PLOTS